Relative Clause Attachment in Jabberwocky and Syntactic Prose

Frank Wijnen
Utrecht University

Fodor's (1998, 2002) implicit prosody hypothesis states that readers, in processing a sentence, create a phonological representation, which can have an impact on syntactic parsing decisions, notably in cases of structural ambiguity. Fodor suggests that the phonological processor divides the input into minimal processing units, or 'chunks', and that this chunking is driven by prosodic structure. Several studies have provided evidence for the hypothesis, but the results of a number of these cannot distinguish between implicit prosody as a last resort (or repair) strategy, or as a factor contributing to first pass parsing. Moreover, it is sometimes difficult to rule out possible semantic and pragmatic effects on chunking. In the present study, we address these issues by looking at globally ambiguous materials (no re-analysis), in which lexical-semantic cues were suppressed: Jabberwocky (functional elements, including the relative pronoun die, are real Dutch, everything else is phonologically legal nonsense, e.g. De kalambulo van de fup die verstritst was.), and syntactic prose (e.g. De grond van de carriere die gebeld werd. 'The soil of the career that was rung.') We constructed 32 NP1-prep-NP2-RC structures for both conditions, and varied the length (weight) of the NPs, as well as that of the RC. In an
on-line questionnaire, we asked subjects to indicate whether they felt the RC modified the first or the second NP. The results of the Jabberwocky condition show an overall preference for low attachment, which is stronger for short RCs (78%) than for long RCs (63%). This supports Fodor's (1998) 'antigravity law'. In items with short RCs, where NP1 is long and NP2 is short, NP2 attachment is markedly more preferred than in constructions in which the relative weights are reversed (84% vs. 69%). When RCs are long, this effect is suppressed. The results of the syntactic prose condition show
the same pattern as those of the Jabberwocky condition, but are less pronounced. Taken together, the results suggest very strongly that phonological structure, apart from lexical, semantic or pragmatic factors, affects first pass parsing. We are currently collecting on-line (SPR) data to corroborate this conclusion.

References

Fodor, J.D. (1998). Learning to parse? Journal of Psycholinguistic Research 27, 285-319.

Fodor, J.D. (2002). Prosodic disambiguation in silent reading. Proceedings of NELS 32.